1999-01-1288

Driver Crash Avoidance Behavior with ABS in an IntersectionIncursion Scenario on Dry Versus Wet Pavement

Elizabeth N. Mazzae and Frank BarickmanNational Highway Traffic Safety Administration

G. H. Scott Baldwin and Garrick ForkenbrockTransportation Research Center Inc.

Copyright © 1999 Society of Automotive Engineers, Inc.

ABSTRACT

The National Highway Traffic Safety Administration(NHTSA) has developed its Light Vehicle Antilock BrakeSystems (ABS) Research Program in an effort todetermine the cause(s) of the apparent increase in fatalsingle-vehicle run-off-road crashes as vehicles undergo atransition from conventional brakes to ABS. As part of thisprogram, NHTSA conducted research examining drivercrash avoidance behavior and the effects of ABS ondrivers’ ability to avoid a collision in a crash-imminentsituation. The study described here was conducted on atest track under dry and wet pavement conditions toexamine the effects of ABS versus conventional brakes,ABS brake pedal feedback level, and ABS instruction ondriver behavior and crash avoidance performance. Thisstudy found that drivers do tend to brake and steer inrealistic crash avoidance situations and that excessivesteering can occur. However, a significant number of roaddepartures did not result from this behavior for eitherpavement condition. ABS was found to reduce crashessignificantly on wet pavement as compared to conventionalbrakes.

INTRODUCTION

Since 1985, antilock brake systems (ABS) have beenincreasingly available on many passenger car and lighttruck make/models. ABS have been sold as an addedsafety feature which enhances drivers’ ability to control avehicle and, in some cases, improves vehicle stoppingperformance. In the interest of reaping the benefits ofABS in terms of a reduction in crashes, the HighwaySafety Act of 1991, Section 2507 charged the NationalHighway Traffic Safety Administration (NHTSA) with thetask of determining whether ABS should be required on allpassenger vehicles.

As a result, NHTSA undertook a series of investigations todetermine the potential benefits of ABS and the effect ofABS on crash rates. Test programs have shown that ABSappear to be very promising safety devices whenevaluated on a test track. Under many pavementconditions antilock brake systems allow the driver to stop

a vehicle more rapidly while maintaining steering controleven during situations of extreme, panic braking. Brakeexperts anticipated that the introduction of ABS onpassenger vehicles would reduce both the number andseverity of crashes. However, a number of crash dataanalyses have been performed in recent years by NHTSA,automotive manufacturers, and others which indicate thatthe introduction of ABS has not been found to beassociated with a reduction in crashes to the expectedextent.

CRASH DATA

Kahane [1] found that, with the introduction of ABS,involvements in multi-vehicle crashes involving fatalities onwet roads were significantly reduced by 24 percent, andnonfatal crashes by 14 percent. However, thesereductions were offset by a statistically significant increasein the frequency of single-vehicle, run-off-road crashes, ascompared to cars without ABS. Run-off-road crashes, asconsidered in this report, included rollovers, side impactswith fixed objects , and frontal impacts with fixed objects.Fatal run-off-road crashes were up by 28 percent andnonfatal crashes by 19 percent. On wet roads, fatal run-off-road crashes increased 17 percent and non-fatal run-off-road crashes increased by 24 percent. On dry roads,fatal run-off-road crashed increased by 29 percent whilenon-fatal crashes increased by 17 percent.

Hertz, Hilton, and Johnson [2] presented results forpassenger car run-off-road crashes according to thefollowing crash types: rollovers, side impacts with parkedvehicles or fixed objects, and frontal impacts with parkedvehicles or fixed objects. For dry roads, ABS was found tobe associated with a 17 percent decrease in rollovercrashes, a 13 percent decrease in frontal impacts withparked vehicles or fixed objects, and a 7 percent increasein side impacts with parked cars or fixed objects. Forpedestrian crashes, ABS was associated with a 30 percentreduction on dry roads and a 10 percent reduction inunfavorable road conditions (i.e., wet, snowy, icy, gravel).In regards to only those crashes involving fatalities, ABSwas found to be associated with a 51 percent increase infatal rollover crashes on dry roads. For fatal side impact

crashes, ABS produced a 69 percent increase forunfavorable road conditions, and a 61 percent increase forfavorable road conditions.

NHTSA’S LIGHT VEHICLE ABS RESEARCH PROGRAM

In an effort to investigate possible causes contributing tothe observed increase in fatal single-vehicle crashesassociated with ABS implementation, NHTSA developedits Light Vehicle ABS Research Program. This programcontains nine separate tasks addressing issues relating topassenger cars and light trucks such as ABS hardwareperformance, examination of ABS crash reports, andassessment of driver behavior with ABS (as outlined in[3]). The cumulative results of these varied tasks willprovide insight regarding ABS effectiveness. To date,NHTSA research has found no systematic hardwaredeficiencies in its examination of ABS hardwareperformance (as documented in [4]) except for thepreviously identified increase in stopping distance ongravel. It is unknown, however, to what extent theincrease in run-off-road crashes may be due to drivers’incorrect usage of ABS, incorrect response to ABSactivation, incorrect instinctive driver response (e.g.,oversteering), changes in driver behavior (e.g., behavioraladaptation) as a result of ABS use, or some other factor.Task 5 of this program, which this paper focuses on,addresses these driver-related issues.

TASK 5: HUMAN FACTORS STUDIES OF DRIVERCRASH AVOIDANCE BEHAVIOR

To determine whether some aspect of driver behavior in acrash-imminent situation may be counteracting thepotential benefits of ABS, NHTSA embarked on a series ofhuman factors studies. These studies, which composeTask 5 of the research program, focus on the examinationof driver crash avoidance behavior as a function of brakesystem and various other factors.

One of the theories Task 5 sought to address was whetherthe apparent increase in fatal single-vehicle crashesinvolving ABS-equipped vehicles may be due tocharacteristics of driver steering and braking behavior incrash-imminent situations. According to this theory, insituations of extreme, panic braking, drivers may have atendency to brake hard and make large steering inputs toavoid a crash. Without four-wheel ABS, aggressivebraking may lock the front wheels of the vehicle,eliminating directional control capability, rendering thedriver’s steering behavior irrelevant. With four-wheel ABS,the vehicle’s wheels do not lock, therefore, the vehicledoes not lose directional control capability during hardbraking and drivers’ steering inputs are then effective indirecting the vehicle’s motion. This directional controlcould result in drivers avoiding multi-vehicle crashes bydriving off the road and experiencing single-vehiclecrashes.

To investigate this theory, Task 5 sought to address issuessuch as whether:

& Drivers tend to both brake and steer (as opposedto only braking or only steering) during crashavoidance maneuvers;

& Drivers tend to make large, potentially excessive,steering inputs during crash avoidancemaneuvers;

& Drivers’ crash avoidance maneuvers in ABS-equipped vehicles result in road departures moreoften than in conventionally braked vehicles;

& Drivers avoid more crashes in ABS-equippedvehicles than in conventionally braked vehicles ondry pavement; and

& Drivers avoid more crashes in ABS-equippedvehicles than in conventionally braked vehicles onwet pavement.

Task 5 of NHTSA’s Light Vehicle ABS Research Programincludes three studies. Two studies were conducted on atest track (one on dry pavement, one on wet pavement)and one on the Iowa Driving Simulator (IDS).

These studies used a right-side intersection incursionscenario to elicit a crash avoidance response from humansubjects. This scenario was chosen because it was likelyto induce steering behavior and had the potential forsubjects driving the vehicle off of the road. This obstacleavoidance scenario is not responsible for all run-off-roadcrashes and results may not be representative of driverbehavior in all situations leading to vehicle road departure.Many run-off-road crashes occur when drivers are unableto maneuver through a curve in the roadway or when theyare drowsy or under the influence of alcohol. However, itis believed that the results of this study will be useful indetermining not only the extent to which drivers are able tomaneuver a vehicle, but also drivers’ physical capacity tosupply control inputs to the vehicle. Insight into drivers’ability to maintain vehicle control during a panic maneuverand ability to avoid a collision can also be gained from thisresearch.

Although the same scenario was involved in each of theseexperimental venues, advantages to both test track andsimulator means of observing driver behavior werepresent. The test track experiments allowed driverbehavior to be examined in a realistic environment atmoderate speeds in real vehicles with simulated obstacleson both dry and wet pavement. The IDS study allowed fordriver behavior to be examined using a highly repeatabletest method in a simulated environment at higher travelspeeds and with no chance of actual physical collision orinjury. This paper discusses the method used and aportion of the results from both the dry and wet pavementtest track studies.

METHOD

SUBJECTS

Subjects for both the dry and the wet pavement studieswere recruited from the central Ohio area using newspaperadvertisements and flyers posted in local commercialestablishments. Interested persons responded to the adsby telephone and were asked a series of questionsregarding health and driving habits to ensure that theywere fit to participate.

All subjects were between 25 and 55 years of age. Eligiblesubjects had no reported health problems which couldadversely affect driving ability. Subjects were required tohold a valid driver’s license, have driven at least 3,000miles in the past year, and be able to drive an automatictransmission vehicle without special equipment.

Subjects who were accepted for participation in thesestudies did not use a vehicle equipped with ABS as theirprimary mode of transportation. In addition, subjects mayhave driven an ABS-equipped vehicle before, but hadnever personally activated ABS.

In order to complete these studies in an economicalmanner and within the required time frame of the researchprogram, fewer subjects were involved in the wetpavement study than in the dry pavement study. The drypavement study used 192 subjects while the wet pavementstudy used 53 subjects.

Subjects were recruited without regard for their occupationand were assumed to be representative of the populationof average drivers. A subsequent examination of subjectdemographics shows that 7 percent were professionaltruck drivers and as many as 4 percent of subjects in thedry pavement study held an occupation that involveddriving in some way. Likewise, approximately 2 percent ofsubjects (1 subject) in the wet pavement study werereported to be truck drivers and another approximately 2percent (1 subject) held an occupation that involved drivingin some way.

EXPERIMENTAL DESIGN

Independent variables in these studies included type ofbrake system, ABS brake pedal feedback level, and ABSinstruction, braking practice, speed limit, pavementcondition, test vehicle, time-to-intersection, and gender.The subset of independent variables common to both thewet and dry test track studies included brake system(ABS, conventional brakes), ABS brake pedal feedbacklevel (light, heavy), and gender. This paper focuses oncomparing the results from each of the two studies forthese three factors in order to draw conclusions regardingdifferences between dry and wet pavement findings.

Both studies involved at least 8 subjects per condition fora total of 192 for the dry pavement study and 53 for thewet pavement study. The order of presentation ofconditions was randomized. Gender was approximatelybalanced per experimental condition.

Brake System, ABS Brake Pedal Feedback, and TestVehicles

Brake system conditions included: a) conventional brakesystem, b) ABS with minimal pedal feedback; and c) ABSwith a large degree of pedal feedback. The use of two testvehicles was required in order to obtain the two ABS brakepedal feedback levels. A 1995 Chevrolet Lumina equippedwith Delco VI ABS represented the light feedbackcondition. A 1996 Ford Taurus equipped with Bosch ABSrepresented the heavy ABS brake pedal feedbackcondition. In order to account for any potential vehicleeffects, both vehicles were also tested in the conventionalcase. To create the conventional brake system condition,the ABS were electronically disabled. A secondaryindependent variable, vehicle, was then examined toidentify any potential confounding effects of vehiclemake/model.

ABS Instruction

To address whether drivers may be more likely to crash inan ABS-equipped vehicle due to lack of knowledge aboutABS, ABS instruction was included as an independentvariable in these studies. Of the subjects receiving theABS condition, half received ABS instruction and the otherhalf received no ABS instruction. ABS instructionconsisted of a short video containing an initial segmentdescribing the use of seat belts, air bag operation, andsafety precautions, as well as a latter segment (taken froman OEM video [5] designed to be given to a buyer with thepurchase of a new vehicle) which illustrated ABS operationand use. Subjects in the conventional brake systemcondition were given no instruction other than the recordedaudio instructions which all subjects received instructingthem how to drive on the test route and test procedures.

Time-To-Intersection

Time-To-Intersection (TTI) was defined as the time itwould take a subject to reach the intersection at theircurrent velocity as measured at a defined "trigger" point inthe roadway. The purpose of this independent variablewas to examine whether subjects altered their collisionavoidance strategy based on the time available to respondto the event. Pilot testing was conducted prior to the maintest to determine and confirm the TTI values to be used.These values were selected to promote driver steering andto represent two conditions: one in which most but not alldrivers would be able to avoid a collision in a vehicleequipped with conventional brakes, and one in which onlyvery few drivers could avoid a crash in a conventionalbrake system equipped vehicle. Due to difficulties withaltering test equipment for accommodation of two TTI

Figure 1 . Illustration of location and layout of course used in the dry and wet pavement intersection incursion studies.

values, only one value (2.5 seconds) was used in the wetpavement study. Results presented in this paper for thedry test track study represent both TTI values combined.

Speed Limit

For safety reasons, speed limits in the test track studieswere kept to 45 mph (72 kph) on dry pavement and 35mph (56 kph) on wet pavement. Results for the 45 mphcondition could be compared to results for the Iowa DrivingSimulator study for the same speed.

Pavement Condition: Dry Versus Wet

The dry test track study was conducted on asphaltpavement having an approximate coefficient of friction of0.9. For this test a simulated intersection (A) wasintegrated into a figure eight shaped course, as shown inFigure 1 [6].

For the wet test track study, a simulated intersection wasconstructed on a Jennite pad which was wetted for testing.In order to accommodate the different location of theintersection (B) for this pavement condition, an oval coursewas created by using only half of the figure eight course,also illustrated in Figure 1. The approximate coefficient offriction of the wet Jennite surface was 0.4.

INSTRUMENTATION

The test vehicles were instrumented with the DataAcquisition System for Crash Avoidance Research(DASCAR). DASCAR is an unobtrusive data acquisitionplatform developed by NHTSA and Oak Ridge NationalLaboratory. This system monitors driver behavior andperformance, vehicle performance, and their associationswith the external environment [7]. The system wasconfigured to record a variety of parameters to describethe dynamics of the vehicle as well as the subjects’ vehiclecontrol inputs. These measures included displacement

and rate of the steering inputs, force applied to the brakepedal, displacement of the throttle, vehicle ground speed,individual wheel speeds, traveled distance, and individualbrake line pressures. These parameters were sampled at200 Hz.

Video cameras were used to collect data both inside andoutside of the vehicle. Within the DASCAR, views wererecorded to collect data on the subjects eye glancebehavior, steering inputs, driver hand position, and throttleand brake applications. A fourth view was used to recordthe forward road scene. These four views weremultiplexed into one video signal using a quad pictureprocessor and simultaneously recorded in synchronizationwith the other measures collected. Additional full-framevideo data was collected by two external sources. Theseviews were in front of and behind the intersection scenarioto capture the test vehicle during the obstacle avoidancemaneuver.

The parameters determined from the data collected usingDASCAR included subject vehicle travel speed atintersection approach and at collision (if a collisionoccurred). Since TTI was based upon an assumed subjectvehicle speed of 45 mph, actual TTI was calculated basedon the test vehicle’s actual speed in case it varied from thedesired value. The time from initiation of incursion vehiclemotion to each of several driver actions was noted, e.g.,throttle release, initial brake application, and initial steeringinput. The magnitude of each of these measures was alsodetermined. Maximum applied brake pedal force,individual wheel speeds, handwheel input, handwheel rate,pitch, and yaw rate were noted. Longitudinal and lateralaccelerations were also recorded. Incidences of wheellockup and ABS activation were also noted for eachsubject. These measures provided a comprehensivequantitative description of drivers’ actions during the crashavoidance maneuver.

A

B

TTI

A = Subject vehicleB = Incurring vehicle (Neon)C = Scenario Vehicle (Saturn)TTI = Time-to-Intersection

C

Figure 2. Illustration of final resting position of incursionvehicle.

PROCEDURE

Intersection Incursion Scenario Implementation

A course composed of a two-lane road was created on theVehicle Dynamics Area (VDA) at the TransportationResearch Center Inc. (TRC) in East Liberty, Ohio. TheVDA is an 1800 foot x 1200 foot, 50 acre flat asphaltsurface. This surface allowed sufficient space to lay outthe simulated intersections without concern that the testvehicle could unintentionally maneuver onto a surface witha different coefficient of friction or leave the paved testsurface entirely. The "figure eight" course on TRC’s VDAwas taken to simulate a rural two-lane highway with lanes12 feet in width. The full figure eight course was used inthe dry pavement testing. For the wet pavement testing,only half of this course was utilized resulting in a small ovalcourse.

An intersection was created on each course by applyingreflective pavement marking tape to define the details ofthe crossing perpendicular roadway and the remainingintersection layout details according to Ohio Department ofTransportation Office of Traffic Engineering specificationsappropriate for the type of roadway being simulated. Toenhance the realism of the intersection, stop lines andcollapsible stop signs were used. Vinyl pylons were placed6 feet from the outer lane lines to indicate where anunpaved off-road surface might begin.

Two vehicles were positioned at the intersection at thestop lines of the crossing lane, a blue 1995 Dodge Neoncoupe on the right and a silver 1992 Saturn SC1 on theleft. To examine subjects’ behavior in response to anunexpected intersection incursion, the car at the right-sidecrossing lane was to be projected 6 feet into the subject’slane of travel. The decision to use only a right-side partialincursion scenario, rather than both full and partialincursions from the left and right sides, was made basedon the results of a previous IDS study [6] and to minimizetest cost.

For safety reasons, the actual vehicles were replaced withrealistic artificial vehicles constructed of polystyrene foamprior to the presentation of the incursion scenario. Imagesof the actual vehicles were silk-screened to the foam andcut out to match the profile of the car. To enable thesesimulated vehicles to stand upright, small trusses alsoconstructed of foam were used.

No provision for longitudinal motion was required for thevehicle on the left side of the intersection since it was notinvolved in the incursion. To secure this stationarysimulated vehicle, two hinges were attached to its wheelsand the VDA. This design allowed it to rest on its printedface and be swung up to its vertical position with ease.

The vehicle on the right side of the intersection wasrequired to allow longitudinal motion in order for it to bemoved into the intersection. To reduce friction between

this movable simulated vehicle and the road surface,Delrin skid plates were used between the foam and thepavement to allow smooth movement of the simulatedvehicle.

Projection of the simulated incursion vehicle (vehicle B)into the intersection was accomplished by attaching a thincable from the front of the simulated incursion vehicle tothe rear of a tow vehicle. The tow vehicle was located farenough from the intersection to prevent it from attractingsubjects’ attention. The simulated vehicle was equippedwith eye screws on the bottom of the trusses which couldbe hooked to two cables anchored to the surface of theVDA. When the subject vehicle drove over a pressure-sensitive tape switch positioned at an appropriate distancefrom the intersection to create the 2.5 or 3.0 second TTI,a display near the tow vehicle was illuminated, alerting itsdriver to pull the foam car into the intersection. When thetow vehicle was driven forward, the simulated vehicle wastowed into the intersection to the specified degree bysliding along the cables until it reached stops attached tothe cables. The foam car was towed 6 feet into thesubject vehicle’s lane of travel to yield a partial incursionas shown in Figure 2.

Ruse

In order to obtain realistic, unbiased driver responses tothe crash imminent scenario presented it was imperativeto ensure that subjects would not perceive that the truepurpose of the study was related to driver behavior in acrash avoidance situation or brake system issues.Experimenters created a ruse to prevent subjects fromanticipating that they would be involved in this crashavoidance exercise. Subjects were told that they wereparticipating in a study of driver behavior in which datawould be collected to assess how average drivers steerand maintain speed while driving in typical driving

conditions. A high technology device, described below,was also introduced for their use part-way into the test tooccupy their attention. Subjects were told that their taskwas to drive normally and that they would be given aquestionnaire to collect information regarding theirimpressions of the drive and use of the high technologydevice.

To help ensure that subjects would not anticipate theintersection incursion event, subjects were informed theywould be driving for approximately 30 minutes. Inactuality, the drive was approximately 15 minutes in length.

Test Procedure

When subjects arrived for testing, they were provided withinformation regarding test procedures and were requiredto sign an informed consent form. Before beginning thetest, some subjects received ABS instruction consisting ofa video tape highlighting the function, behavior, and use ofan antilock brake system. This video also containedsegments addressing seat belt usage and air bag functionin an effort to disguise the focus of the study. Providingcomparable video instruction for the conventional brakecondition was desired to prevent confounding of the data,however, no instruction was found to be readily availablefor this purpose.

Upon entering the test vehicle, subjects were required tolisten to audio instructions which were recorded on acompact disc (CD) which was played by the experimenter.An initial track was played to describe the overall test.Later, other separate CD tracks were played throughoutthe test to describe braking practice procedures (ifapplicable) and the use of a high technology device whichwas used as a distractor task.

Before starting on the test route, some subjectsparticipating in the dry test track study received brakingpractice consisting of 3 brake stops on a wet Jennite-paved surface. Although the effect of practice with ABSon success in avoiding a crash was of primary interest,braking practice was provided to a portion of subjects ineach brake condition in order to prevent confounding ofdata by giving those driving ABS-equipped test vehiclesmore familiarity with the vehicle before experiencing theincursion scenario. This practice gave subjects driving anABS-equipped vehicle the opportunity to experience thebrake pedal feedback present in current ABS.

At all times when a subject was in the test vehicle, anexperimenter was present in the back seat to direct themthrough the test route. Subjects were instructed by theexperimenter to drive on the specified course. A "lead"vehicle operated by a professional driver was scripted todrive by on the course in front of the subject vehicle at theprecise moment that the subject was ready to start ontothe course. As the lead vehicle passed, the subject wastold, "There’s another subject in the study just like you;please turn onto the course behind them and begin your

drive." The purpose of this vehicle was to help encouragesubjects to believe that if the lead vehicle made it throughthe intersection without incident that the subject vehiclewould do the same. In addition, this lead vehicleencouraged subjects to maintain the specified speed limit.

Each subject completed 3.5 laps of the course. After thefirst lap, subjects were instructed to begin using a hightechnology "Laser Rangefinder" device which was installedon the test vehicle. This device consisted of a lasermounted in the grill of the test vehicle which detected thedistance to a forward vehicle. A display was mounted atthe center of the dashboard which related the distanceinformation. Subjects were told to use the informationprovided by the display in order to maintain a distance of200 feet from the forward vehicle which was traveling atthe specified speed limit for that study. Use of the systemprovided a distraction for subjects, helped to prevent themfrom realizing the true aim of the test, and also helpedthem maintain the desired travel speed.

As the lead and subject vehicles passed through theintersection the first two times, the actual scenario vehicleswere positioned at the intersection. Between the secondand third laps, drivers entered the two vehicles when thesubject vehicle was out of sight so that on the third lap, thesubject could see drivers in the scenario vehicles.Between the third and fourth laps, the artificial vehicleswere set in place. The driver of the silver Saturn exited thecar, lifted the Saturn likeness into place, and latched it tothe VDA. At the same time the driver and back seatoccupant of the Neon quickly exited their car and attachedthe simulated incursion vehicle to its cables. The realscenario vehicles were then removed to a remote locationout of view of the subject. When the subject passed overthe tape switch on the fourth lap, the simulated vehiclewas towed into the lane and stopped with the front of thevehicle 6 feet into the subject’s lane of travel. Followingthis event the experiment ended.

RESULTS

Crash Avoidance Strategy - Overall

In the wet pavement study, 98 percent of the 53 subjectsattempted both steering and braking inputs during theavoidance attempt. Of the 98 percent that braked andsteered, 46 percent applied the brakes before steering, astheir initial response, while 52 percent steered beforebraking. A summary of these results are presented inTable 1.

In the dry pavement study, 94 percent of the192 subjectsattempted both steering and braking inputs in an effort toavoid colliding with the incursion vehicle. Of this 94%, 47percent applied the brakes before steering, as their initialresponse, while 46 percent steered before braking.

63

76

31

18

68

75

30

23

Steering Maneuver

Initial LeftAvoid Left

Initial RightAvoid Right

0

10

20

30

40

50

60

70

80

Dry Wet

Figure 3. Steering response directions during incursionscenario.

Driver Input Dry Wet

Braked and steered 94 98

- Braked then steered 47 46

- Steered then braked 46 52

Braked only 6 2

Steered only 0 0

Table 1. Percent of subjects who braked and/or steeredduring the intersection incursion scenario.

For the initial steering maneuver (defined as the firststeering input of magnitude greater than six degrees whichthe subject made after the initiation of the incursionvehicle’s motion), 63 percent of the 192 subjects chose tosteer left and 31 percent chose to steer right (12 subjectsdid not steer) in the dry pavement study. In the wetpavement study, 68 percent of the 53 subjects chose tosteer left and 30 percent chose to steer right (one subjectdid not steer).

An "avoidance steering input" was defined as the steeringinput which the subject intended to cause the subjectvehicle to travel around the foam vehicle without collidingwith it; i.e., the steering input which was in progress as thesubject vehicle passed through the plane of motion of theincursion vehicle. Differentiating between initial steeringinput and avoidance steering input provides someinformation regarding how many subjects changed thedirection of their steering input while trying to avoid a crashwith the incursion vehicle (see Figure 3). Seventy-sixpercent of the 192 subjects chose to try to steer left of theencroaching vehicle and 18 percent made the decision tosteer right as an avoidance steering input on dry

pavement. For the wet pavement study, 75 percent of the53 subjects chose to try to steer left of the encroachingvehicle and 23 percent made the decision to steer right toavoid a collision. In the dry pavement study, 36 percent ofthose whose avoidance steering input was toward the leftcrashed, while 47 percent who steered right crashed (8 of12 subjects crashed who did not steer at all). In the wetpavement study, 70 percent of the subjects who steeredleft in their avoidance steering input crashed, while 75percent who steered right crashed. The wet pavementstudy subject who did not steer also crashed.

Steering Behavior - Overall

The average magnitude of avoidance steering inputobserved was 53 degrees for the dry pavement study and74 degrees for the wet pavement study. The highestobserved steering input from an individual subject duringthe avoidance maneuver in the dry pavement study was271 degrees, while for the wet pavement study, this valuewas 289 degrees.

The average maximum steering rate obtained during theavoidance maneuver was 262 degrees per second for thedry pavement study and 294 degrees per second for thewet pavement study. The highest observed steering rateachieved by a subject in the dry pavement study was 1159degrees per second, and was 1335 degrees per second inthe wet pavement study. Ninety-five percent of steeringrates observed were less than 600 degrees per second inthe dry pavement study and less than 643 degrees persecond in the wet pavement study.

Braking Behavior - Overall

The overall average maximum brake pedal forces obtainedwere 64 pounds and 68 pounds for the dry and wetpavement studies, respectively. The highest observedbrake pedal force input generated by a subject in the drypavement study was 188 pounds. In the wet pavementstudy, the highest observed brake pedal force inputgenerated by a subject was 240 pounds.

On dry pavement, 31 percent of the 192 subjects eitheractivated ABS or locked the vehicle’s wheels withconventional brakes during the avoidance maneuver. Inthe wet pavement study, 96 percent of the 53 subjectseither activated ABS or locked the vehicle’s wheels withconventional brakes during the avoidance maneuver.

Road Departures - Overall

In the dry pavement study, two people out of 192 fullydeparted the roadway during the collision avoidancemaneuver. One subject steered left around the front of theencroaching vehicle and ran off the road to the left. Theother subject departed the road to the right after hitting therear of the incursion vehicle. No four-wheel roaddepartures were seen in the wet pavement study.

5361

49

74

103

61

Avoidance Steering Magnitudes

Overall Conventional ABS

0

20

40

60

80

100

120

Dry Wet

Figure 4. Magnitude of avoidance steering inputs.

Figure 5. Avoidance steering rates.

Crashes - Overall

During the intersection incursion event, 40 percent of the192 subjects collided with the incursion vehicle on drypavement and 72 percent of the 53 subjects crashed onwet pavement.

BRAKE SYSTEM: ABS VS. CONVENTIONAL

Crash Avoidance Strategy by Brake System

For those subjects who braked then steered during thecrash avoidance maneuver, the delay time from when theyinitiated braking to when they began to steer did not differsignificantly by brake system. Subjects in the conventionalbrake system condition waited 0.5 seconds after brakingbefore they initiated steering, while those with ABS waited0.55 seconds.

Steering Behavior by Brake System

On dry pavement, the average magnitudes of theavoidance steering input was 49 degrees for those withABS and 61 degrees for those driving a vehicle equippedwith a conventional brake system, as shown in Table 2 andFigure 4. However, this difference was not significant [p =0.4535]. On wet pavement, the average magnitude of theavoidance steering input was significantly different [p =0.0095] with ABS subjects having inputs averaging 61degrees and conventional brake system subjects at 103degrees.

As shown in Table 2 and Figure 5, on dry pavement, theaverage maximum steering rate of the avoidancemaneuver was 246 degrees per second for ABS subjectsand 296 degrees per second for the conventional brakesystem case [p = 0.5701]. In the wet pavement study, theaverage maximum steering rate of the avoidancemaneuver for ABS subjects was 313 degrees per second,while for the conventional group this value was 255degrees per second [p = 0.8649].

Table 2 also lists results for the time from initiation ofincursion motion until subjects applied their largest steeringinput. In the wet pavement study, for subjects in the ABScondition, the time to maximum steering input wassignificantly less [p = 0.0003], 2.39 seconds, than thatobserved for conventional brakes (3.37 seconds). On drypavement, the time to maximum steering input for the ABSsubjects was quite similar to the conventional brakesubjects, at 2.73 seconds and 2.70 seconds, respectively.

Steering InputCharacteristics

BrakeSystem

Dry Wet

Average magnitude ofsteering input(degrees)

Conventional 61 103*

ABS 49 61*

Average maximumsteering input rate(degrees per second)

Conventional 296 255

ABS 246 313

Time to maximumsteering input(seconds)

Conventional 2.70 3.37*

ABS 2.73 2.39*

Table 2. Characteristics of observed steering behavior bybrake system and pavement condition. (* These values arestatistically significantly different.)

64 6265

68

59

72

Brake Force

Overall Conventional ABS

0

10

20

30

40

50

60

70

80

Dry Wet

Figure 6. Brake pedal forces.

Braking Behavior by Brake System

As shown in Figure 6, the average maximum brake pedalforce observed for subjects in the dry pavement study ABScondition during the avoidance maneuver was 65 pounds.For dry pavement subjects in the conventional brakesystem case, the average was similar at 62 pounds. Thisdifference was not statistically significant [p = 0.8852].The wet pavement study did not have significantdifferences either [p = 0.4730], with ABS at 72 pounds andconventional at 59 pounds.

One might expect that observed brake pedal applicationdurations should be longer for ABS if drivers were usingthe ABS properly, i.e., maintaining pressure on the brakepedal and not "pumping" the pedal. The average brakepedal application duration observed during the crashavoidance maneuver in the dry pavement study was in factlonger for ABS (2.20 seconds) than for conventional (1.99seconds); however, this difference was not statisticallysignificant. In the wet pavement study, the opposite resultwas found. The brake pedal application duration for ABS(2.08 seconds) was shorter than for conventional (3.23seconds) [p = 0.0864](see Table 3).

Based on a cursory examination of the video recordingsof each subject’s experience of the intersection incursionscenario from both the dry and wet pavement experiments,no evidence was observed of subjects either pulling theirfoot off of the brake pedal due to being startled by ABSpedal feedback or attempting to "pump" the brake pedalwith ABS.

One also might expect that subjects receiving ABSinstruction might have longer brake pedal applicationdurations as a result of being told not to "pump" the brakepedal with ABS. This was true in the dry pavement study.Subjects receiving ABS instruction had an average brake

pedal application duration (2.56 seconds) which wassignificantly longer than for those with ABS who receivedno instruction (1.91 seconds) [ p = 0.0220]. However, inthe wet pavement study, subjects receiving ABSinstruction had an average brake pedal applicationduration (2.09 seconds) which was similar to those withABS who received no instruction (2.06 seconds).

Braking InputCharacteristics

BrakeSystem

Dry Wet

Average maximumbrake pedal force(pounds)

Conventional 62 59

ABS 65 72

Brake pedalapplication duration(seconds)

Conventional 1.99 3.23

ABS 2.20 2.08

ABS (noinstruction)

1.91* 2.06

ABS(instruction)

2.56* 2.09

Table 3. Characteristics of observed braking behavior bybrake system and pavement condition. (* These valuesare statistically significantly different.)

Road Departures by Brake System

In the dry pavement study, 2 subjects out of 192 drovecompletely off the road (all four wheels) during theavoidance maneuver. Both of these subjects whoexperienced road departures were in the ABS condition (of128 subjects) and both activated ABS during their crashavoidance maneuver. None of the 64 subjects who hadconventional brakes drove completely off the road duringthe avoidance maneuver. In the wet pavement study, noone drove completely off the road.

In addition, two partial (two-wheel) road departures werealso observed in the dry pavement study. One of thecases involved a subject driving with ABS while the otherinvolved conventional brakes. In the wet pavement study,only one subject had one wheel which crossed over theedge line. That subject was driving with conventionalbrakes.

Crashes by Brake System

Table 4 summarizes results for subjects ability to avoid acrash according to brake system and pavement condition,as measured by number of crashes. In the dry pavementstudy, fewer subjects crashed into the encroaching vehiclein the ABS condition (38 percent of 128), than in theconventional brake system condition (45 percent of 64).However, this difference was not statistically significant.In the wet pavement study, 100 percent of the 17 subjects

246 246 246

313

362

264

Avoidance Steering Rates

Both Light Heavy

0

50

100

150

200

250

300

350

400

Dry Wet

Figure 8 . Avoidance steering rates by brake pedalfeedback.

Figure 7 . Magnitude of avoidance steering inputs bybrake pedal feedback.

in the conventional brake system condition crashed intothe encroaching vehicle and only 58 percent of the 36subjects in the ABS condition crashed. This differencewas statistically significant [p = 0.0020].

Brake System Dry Wet

Conventional 45 100*

ABS 38 58*

ABS (light brake pedalfeedback)

33 72

ABS (heavy brake pedalfeedback)

42 44

Table 4. Percent of subjects who crashed in to theincursion vehicle as a function of brake system andpavement condition. (* These values are statisticallysignificantly different.)

ABS BRAKE PEDAL FEEDBACK

Steering Behavior by ABS Brake Pedal Feedback

In the dry pavement study, the average avoidance steeringinput magnitude observed was 49 degrees for ABSsubjects in both the light and heavy ABS brake pedalfeedback conditions. On wet pavement, the averagemagnitude of the avoidance steering input was 58 degreesfor ABS subjects in the light feedback condition and 64degrees for the heavy ABS brake pedal feedback condition(see Figure 7).

On dry pavement, the average maximum steering rateobserved during the avoidance maneuver for both light andheavy ABS brake pedal feedback conditions was 246degrees per second, as shown in Figure 8. In the wetpavement study, the average maximum steering rate was362 degrees per second for the light feedback conditionand only 264 degrees per second for the heavy feedbackcondition.

Braking Behavior by ABS Brake Pedal Feedback

In the dry pavement study, subjects in the light feedbackcondition had an average maximum brake pedal force,during the avoidance maneuver, of 64 pounds. Theaverage was similar at 65 pounds for the heavy feedbacksubjects. In the wet pavement study however, the lightfeedback subjects had an average maximum pedal forceof 81 pounds and the heavy feedback subjects averagedlower maximums at 63 pounds of brake pedal force (seeFigure 9).

Road Departures by ABS Brake Pedal Feedback

As stated earlier, two vehicles departed the roadway fullyin the dry pavement study. Both of these road departuresinvolved the heavy brake pedal feedback ABS condition of64 subjects. In addition, two partial (two-wheel) roaddepartures were observed, one with ABS and oneconventional. Each of these cases occurred in the samevehicle (the heavy brake pedal feedback vehicle). In thewet pavement study, no road departures occurred.

65 64 6572

81

63

Brake Force

Both Light Heavy

0

20

40

60

80

100

Dry Wet

Figure 9 . Brake forces by brake pedal feedback.

Figure 10 . Magnitude of the avoidance steering input bygender and brake system - dry study.

Crashes by ABS Brake Pedal Feedback As shown in Table 4, 33 percent of the 64 subjects collidedwith the incursion vehicle on dry pavement while drivingthe light feedback vehicle with ABS. In the heavyfeedback ABS vehicle, 42 percent of those 64 subjectscrashed on dry pavement.

On wet pavement, 72 percent of the 18 subjects drivingthe light feedback ABS vehicle crashed, whereas 44percent of the 18 subjects in the heavy feedback ABScondition crashed on wet pavement. However, this 28percent difference in the crash rates was not statisticallysignificant.

GENDER

Gender was balanced throughout all conditions in thisstudy. Overall, crash avoidance behavior observed formale subjects was characterized by inputs of highermagnitudes than females. However, most of thedifferences observed between genders were notstatistically significant.

Steering Behavior by Gender

Table 5 and Figure 10 show the average magnitudes ofthe avoidance steering input for females in the drypavement study was 45 degrees and for males was 59degrees. The average magnitude of the avoidancesteering input for females in the wet pavement study was72 degrees and for males was 77 degrees (see Figure 11).These differences between genders were not statisticallysignificant.

The average maximum steering rate in any direction forfemales in the dry pavement study was 213 degrees persecond, as listed in Table 5. For males, the averagemaximum steering rate observed on dry pavement was

306 degrees per second. The average maximum steeringrate in any direction for females in the wet pavement studywas 269 degrees per second. For males, the averagemaximum steering rate observed on wet pavement was321 degrees per second.

Steering InputCharacteristics

Gender Dry Wet

Average magnitude ofavoidance steering input(degrees)

Female 45 72

Male 59 77

Average maximumsteering input rate(degrees per second)

Female 213 269

Male 306 321

Table 5. Characteristics of observed steering behavior bygender.

Braking Behavior by Gender

Table 6 summarizes braking behavior observed during theincursion scenario as a function of gender. The averagemaximum brake pedal force was 61 pounds for femalesand 66 pounds for males in the dry pavement study. Theaverage maximum brake pedal force was 62 pounds forfemales and 74 pounds for males in the wet pavementstudy. On dry pavement, females first applied the brakeas a reaction to the incursion vehicle at 1.51 seconds andmales at 1.48 seconds on average. In the wet pavementstudy, females first applied the brake as a reaction to theincursion vehicle at 1.37 seconds and males at 1.47seconds on average. These differences were notstatistically significant.

Figure 11 . Magnitude of the avoidance steering input bygender and brake system - wet study.

40

50

35

40 39 40

Crashes - Dry

MaleMale Conv

Male ABSFemale

Fem ConvFem ABS

0

10

20

30

40

50

Figure 12. Crashes by gender and brake system - drypavement study.

77

100

62.567

100

55

Crashes - Wet

MaleMale Conv

Male ABSFemale

Fem ConvFem ABS

0

20

40

60

80

100

Figure 13. Crashes by gender and brake system - wetstudy.

Braking InputCharacteristics

Gender Dry Wet

Average maximum brakepedal force (pounds)

Female 61 62

Male 66 74

Brake reaction time(seconds)

Female 1.51 1.37

Male 1.48 1.47

Table 6. Characteristics of observed braking behavior bygender and pavement condition.

Road Departures by Gender

Of the two subjects in the dry pavement study who drovecompletely off the road to avoid a crash, the subject whodeparted the road to the right was a male, while thesubject who departed the road to the left was a female. Ofthe two partial, two-wheel road departures observed inthese studies, both occurred in the dry pavement study aswell. The right side partial road departure was made by amale subject and the left side departure was made by afemale.

Crashes by Gender

On dry pavement, 40 percent of 88 females (40% ABS,39% conventional) collided with the incursion vehicle.Forty percent of 104 males (35% ABS, 50% conventional)also crashed in this study (see Figure 12).

In the wet pavement study, 67 percent of 27 females (55%ABS, 100% conventional) collided with the incursion

vehicle while 77 percent of 26 males (63% ABS, 100%conventional) also crashed (see Figure 13).

DISCUSSION

Do drivers tend to both brake and steer in crash-imminentsituations

Nearly all subjects in these studies both braked andsteered in an attempt to avoid colliding with the incursionvehicle. For subjects that both braked and steered in anattempt to avoid a collision, the likelihood of whether theybraked or steered first was approximately equal for bothpavement conditions.

Do people exhibit excessive steering behavior during crashavoidance maneuvers?

In general, steering inputs exhibited by subjects in thesetest track studies were smaller and slower than thoseobserved in the related IDS study [8]. This difference isbelieved to be attributable to the lack of "road feel" presenton the IDS as well as the limited range of travel of thesimulator motion base. However, the observedmagnitudes and rates of steering inputs for these test trackstudies were still relatively large. As noted in the relatedIDS study [8], drivers appeared to alter their steeringbehavior based on the degree to which they felt thesteering inputs were affecting the motion of the vehicle inthe desired direction. Subjects with ABS made smallersteering inputs and used lower steering input applicationrates than subjects with conventional brakes. The reasonfor this is believed to be that subjects made increasinglylarge steering inputs with conventional brakes since, withlocked wheels, their steering inputs were not effective indirecting the vehicle’s motion.

Do people have more road departures in ABS-equippedvehicles than in conventionally brakes vehicles?

Overall, results from this research indicate that, althoughsubjects were observed making steering inputscharacterized by large magnitudes and high rates, theseaggressive steering inputs did not result in a significantnumber of road departures. In addition, ABS was notassociated with significantly higher frequencies of roaddepartures in these studies than were observed in theconventional brake system conditions.

Do people crash less frequently in ABS-equipped vehiclesthan in vehicles equipped with conventional brakes?

On wet pavement, ABS was associated with significantlyfewer crashes than conventional brakes. Although fewersubjects crashed with ABS than with conventional brakeson dry pavement, this difference was not significant.

Also on wet pavement, 28 percent fewer crashes wereobserved with ABS with heavy brake pedal feedback thanwith light feedback ABS, however, this result was notstatistically significant. On dry pavement, more subjectscrashed in the heavy ABS feedback condition than in thelight feedback ABS condition but not at a significant level.Thus, based on these findings as well as the fact that nosubject was observed pulling their foot off of the brakepedal due to being startled by haptic feedback, it is notimmediately clear what level of ABS-related brake pedalfeedback is most beneficial, appropriate, or lends to fewercrashes.

Overall, ABS was not associated with a significantlygreater number of crashes than conventional brakes forany factor discussed in this paper.

CONCLUSIONS

An experiment was conducted in which drivers’ collisionavoidance behavior in a simulated right-side intersectionincursion scenario was examined as a function of vehiclebrake system (conventional, ABS) and pavement condition(dry, wet). Subjects in these studies demonstrated thecapability to make aggressive steering and braking inputs.However, despite the high magnitudes and rates of manyinputs observed, few road departures were observed.Road departures which were observed could not be judgedattributable to ABS performance nor driver interaction withABS. ABS was associated with significantly fewer crasheson wet pavement as compared to conventional brakes.

In conclusion, the results of this study do not appear toindicate that a problem exists due to driver crashavoidance behavior or driver interaction with ABS whichwould contribute to the apparent increase in fatal single-vehicle crashes as identified in conjunction with vehiclestransitioning from conventional to antilock brake systems.Results from this study will be examined in conjunctionwith the results of other tasks included in NHTSA’s LightVehicle ABS Research Program to determine whether thecollective results viewed as a whole provide some insightinto the cause of the increase in fatal single-vehiclecrashes observed in conjunction with the implementationof ABS.

ACKNOWLEDGMENTS

The authors acknowledge the support of W. Riley Garrottof NHTSA for this task and the contributions of Mark Flickof Radlinski and Associates toward the conduct of the drytest track study.

CONTACT

Additional information regarding this research can beobtained by contacting Elizabeth Mazzae, NHTSA VehicleResearch and Test Center, 10820 SR 347, East Liberty,Ohio, 43319 or by e-mail at emazzae@nhtsa.dot.gov.

REFERENCES

1. Kahane, C. J. (December, 1994). "PreliminaryEvaluation of the Effectiveness of Antilock BrakeSystems for Passenger Cars." Technical Report No.808 206. Washington, DC: National Highway TrafficSafety Administration.

2. Hertz, E., Hilton, J., and Johnson, D. M. (1998)."Analysis of the Crash Experience of VehiclesEquipped with Antilock Brake Systems (ABS) - AnUpdate." ESV Paper No. 98-S2-O-07. EnhancedSafety of Vehicles Conference, February, 1998.

3. Garrott, W. R. and Mazzae, E. N. (1999). AnOverview of the National Highway Traffic SafetyAdministration’s Light Vehicle Antilock Brake Systems

Research Program. SAE Paper No. 1999-01-1286.Warrendale, PA: Society of Automotive Engineers.

4. Forkenbrock, G., Flick, Mark, and Garrott, W. G.(1999). A Comprehensive Light Vehicle AntilockBrake System Test Track Performance Evaluation. Prepared for the 1999 SAE International Congressand Exposition. SAE Paper No. 1999-01-1287.Warrendale, PA: Society of Automotive Engineers.

5. General Motors Corporation (1994). Your NewVehicle’s Antilock Brake System; How It Works ForYou. Part No. 15709859.

6. Transportation Research Center Inc. (1998). Partialillustration of TRC Vehicle Dynamics Area. Modifiedfrom complete image obtained at www.trcpg.com.Transportation Research Center Inc., East Liberty,OH.

7. Carter. R. J., Barickman, F. S. (1995). "DataAcquisition System for Crash Avoidance Research."Peer Review of the National Highway Traffic SafetyAdministration Program, pp. 53-63. IntelligentTransportation Society of America.

8. Mazzae, E. N., Baldwin, G. H. S., and McGehee, D. V.(1999). Driver Crash Avoidance Behavior with ABS inan Intersection Incursion Scenario on The IowaDriving Simulator. SAE Paper No. 1999-01-1290.Warrendale, PA: Society of Automotive Engineers.

ADDITIONAL SOURCES

Additional information regarding this work performed underNHTSA’s Light Vehicle ABS Research Program can befound in an upcoming NHTSA report. Status updates onthe progress of the research program are provided on theinternet at the following web site address:www-nrd.nhtsa.dot.gov/vrtc/ca/lvabs.htm.